Solution Structure and Dynamics of LptE from Pseudomonas

Article pubs.acs.org/biochemistry

Solution Structure and Dynamics of LptE from Pseudomonas aeruginosa Kerstin Moehle, Harsha Kocherla, Bernadett Bacsa, Simon Jurt, Katja Zerbe, John A. Robinson,* and Oliver Zerbe* Department of Chemistry, University of Zurich, Winterthurerstrasse 190, CH-8057 Zürich, Switzerland S Supporting Information *

ABSTRACT: LptE is an outer membrane (OM) lipoprotein found in Gramnegative bacteria, where it forms a complex with the β-barrel lipopolysaccharide (LPS) transporter LptD. The LptD/E complex plays a key role in OM biogenesis, by translocating newly synthesized LPS molecules from the periplasm into the external leaflet of the asymmetric OM during cell growth. The LptD/E complex in Pseudomonas aeruginosa (Pa) is a target for macrocyclic β-hairpin-shaped peptidomimetic antibiotics, which inhibit the transport of LPS to the cell surface. So far, the three-dimensional structure of the Pa LptD/E complex and the mode of interaction with these antibiotics are unknown. Here, we report the solution structure of a Pa LptE derivative lacking the N-terminal lipid membrane anchor, determined by multidimensional solution nuclear magnetic resonance (NMR) spectroscopy. The structure reveals a central five-stranded β-sheet against which pack a long C-terminal and a short N-terminal α-helix, as found in homologues of LptE from other Gram-negative bacteria. One unique feature is an extended C-terminal helix in Pa LptE, which in a model of the Pa LptD/E complex appears to be long enough to contact the periplasmic domain of LptD. Chemical shift mapping experiments suggest only weak interactions occur between LptE and the oligosaccharide chains of LPS. The NMR structure of Pa LptE will be valuable for more detailed structural studies of the LptD/E complex from P. aeruginosa.

M

in the periplasm is delivered by LptA to the periplasmic domain of LptD (N-LptD). The lipid chains of LPS are buried in a hydrophobic groove in the β-jelly roll domain (folded antiparallel β-sheets that form a twisted boat structure) present in LptA, LptC, and LptD. While the lipid chains are bound first to LptA and then to N-LptD, the carbohydrate portion of LPS remains exposed to the aqueous compartment in the periplasm. A complex dynamic process then likely ensues in LptD coordinated with the migration of LPS molecules into their final location in the outer leaflet. In this process, a transient opening of terminal β-strands 1 and 26 in the LptD β-barrel forms a lateral gate that allows LPS to fully exit into the OM. Although the LptD/E complex is very stable to dissociation, it is so far unclear whether LptE inside the β-barrel has an active role in the transfer of LPS through the β-barrel. In this work, we focus on LptE (PA3988) from Pseudomonas aeruginosa PAO1. P. aeruginosa (Pa) is a nosocomial pathogen causing high rates of mortality among immunocompromised patients, while also being a frequent cause of declining lung function in cystic fibrosis patients. P. aeruginosa infections can be very difficult to treat, as Pa cells express a diverse set of multidrug efflux pumps, have an especially impermeable OM,

ost Gram-negative bacteria have a cell envelope consisting of an inner membrane (IM), a periplasm containing peptidoglycan, and a unique asymmetric outer membrane (OM) having a phospholipid inner leaflet and an outer leaflet composed almost entirely of lipopolysaccharide (LPS). During growth, LPS is transported from its site of biosynthesis on the IM to the cell surface by eight essential proteins, including the IM flippase MsbA and a small family of LPS transport proteins (LptA−G).1−3 Lipoprotein LptE is found in the outer membrane (OM) in a complex with the βbarrel protein LptD. The LptD/E complex plays a key role in the final step of transport of lipopolysaccharide (LPS) to the cell surface during OM biogenesis (Figure 1A), by translocating LPS molecules from the periplasmic face into the exposed outer leaflet of the asymmetric OM.4 Recent crystal structures of the LptD/E complex from Shigella f lexneri and Salmonella typhimurium show LptE located inside the C-terminal β-barrel domain of LptD (Figure 1B).5,6 LptE is essential in Escherichia coli and is required for the efficient folding of the β-barrel domain of LptD,7,8 mediated by the Bam folding machine.9 Moreover, LptE interacts with LPS in vitro,10,11 suggesting a dual role for LptE both as a structural component of the LptD/ E complex and as a recognition site for LPS at the OM. A model for LPS translocation by LptD/E has been proposed on the basis of extensive crystallographic, photolabeling, mutagenesis, and computational studies.5,6,12−15 In this model, LPS © XXXX American Chemical Society

Received: April 5, 2016 Revised: May 10, 2016

A

DOI: 10.1021/acs.biochem.6b00313 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

Figure 1. (A) LPS transport pathway from the IM to the OM with the proteins LptA−G highlighted. The LptD/E complex in the OM translocates LPS from the periplasm into the outer leaflet. (B) X-ray structure of the LptD/E complex from S. f lexneri (Protein Data Bank entry 4Q35).6 The LptD N-terminal β-jelly roll domain (orange) and the C-terminal β-barrel domains (green and white) are shown, with LptE (light blue) inside the βbarrel. (C) Peptidomimetic antibiotic L27-11.18

0.2−0.45 mM in phosphate buffer [50 mM KPi (pH 6.5) and 25 mM NaCl]. For rLptE concentrations above 0.3 mM, 20 mM CHAPS was added to prevent aggregation. The [15N]-, [15N,13C]-, and [2H,13C,15N]rLptE samples used for measurements were at concentrations of 0.45, 0.35, and 0.4 mM, respectively. Residual dipolar couplings were recorded using [2H,13C,15N]rLptE (0.2 mM) after adding 40 μL of a 50 mg/ mL solution of pf1 phages to a volume of 450 μL. All spectra were recorded at 305 K. LPS used for binding studies was isolated by phenol extraction from Pa PAO1 cells, as described previously.20 Assignments and Relaxation Data. Sequence-specific resonance assignments were obtained using [15N,1H]-HSQC, HNCO , HN(CA)CO, H NCACB, HN(CO)CACB, HBHANH, and HBHA(CO)NH spectra.21−23 Briefly, to obtain backbone assignments, complementary sets of HNCACB/HN(CO)CACB as well as HNCO/HN(CA)CO experiments were conducted using [2H,13C,15N]rLptE. Side chain assignments were obtained from HCCH-TOCSY experiments with 13C mixing times of 14 ms. Initially, amide moieties were picked in the [15N,1H]-HSQC spectrum, and peak positions were confirmed by HNCO correlations. Sequential correlations were obtained by matching CA/CB or CO peaks from the HN(CO)CACB/HNCACB and HNCO/HN(CA)CO pairs. HB and HA correlations were picked in the (H)C(CCO)NH or HBHA(CO)NH spectra, and the resulting HA/CA or HB/CB anchors were used as entries to complete the side chain assignments in the HCCH-TOCSY spectra. Aromatic ring system assignments were obtained from [13 C,1H]-HSQC and 13C-resolved 3D NOESY spectra. Sequential connectivities could be traced with the exception of those of residues 51 and 145, for which resonances were exchange-broadened (vide inf ra). RDCs were measured in pf1 phages as the alignment medium (14 Hz deuterium splitting) using [15N,13C,2H]rLptE. A set of RDCs for HN, CαC′, HC′, and NC′ were measured.24 HN RDCs were derived from IPAP-[15N,1H]-HSQC spectra and CαC′ RDCs from a HNCO experiment, from which the Cα decoupling pulse in t1 was omitted, and the HC′ and NC′ RDCs were extracted from a [15N,1H]-HSQC spectrum in which the general 13C decoupling pulse in t1 was replaced by a Cα-selective inversion pulse. All RDC data were recorded at 600 MHz. The RDCs were extracted using the program suite CCPN that utilizes line-shape deconvolution.

and are acquiring resistance to most currently available classes of antibiotics.16,17 However, an interesting new family of macrocyclic peptidomimetic antibiotics that act specifically against Pseudomonas spp. was reported recently [e.g., L27-11 (shown in Figure 1C)] with a novel mechanism of action, which involves binding to the LptD/E complex and inhibition of the transport of LPS to the cell surface.18,19 An analogue with optimized druglike properties, called POL7080, is now in a phase II clinical study in hospital patients with ventilatorassociated P. aeruginosa pneumonia (National Institutes of Health clinical trial identifier NCT02096328). The unique ability of these peptidomimetic antibiotics to target Pseudomonas spp., but not other Gram-negative microorganisms, derives from the unique three-dimensional (3D) structure of the Pseudomonas LptD/E complex, which is so far unknown. In this work, we set out to study the structure of a recombinant soluble LptE from P. aeruginosa PAO1 by nuclear magnetic resonance (NMR). Using NOE-derived upper distance limits and restraints from residual dipolar couplings (RDCs) measured in a phage alignment medium, we have derived a structure with a 0.4 Å backbone root-mean-square deviation (rmsd) when superimposing well-structured parts of the protein.

■

EXPERIMENTAL PROCEDURES rLptE Production and NMR Sample Preparation. Part of the lptE gene (PA3988) was amplified by polymerase chain reaction (see the Supporting Information for primer sequences) from P. aeruginosa PAO1 chromosomal DNA and cloned SpeI/ XhoI into a modified pET14b vector. This vector includes a 5′ terminus encoding a tobacco etch virus (TEV) protease cleavage site (see the Supporting Information). The rLptE contains three extra non-native N-terminal residues (GTS). Expression of rlptE in E. coli C41(DE3) was induced at an OD600 of ≈0.6 with 0.1 mM isopropyl β-D-1-thiogalactopyranoside, followed by growth at 20 °C for 12 h. For 13C- and 15Nlabeled protein, the expression was performed in M9 medium at 25 °C overnight using [13C]glucose and 15NH4Cl as the sole carbon and nitrogen sources, respectively. rLptE was purified using Ni-NTA Superflow resin. After TEV cleavage, another Ni-NTA column was followed by anion exchange chromatography (MQ10/10). The mass determined by ES-MS of 15Nlabeled rLptE was 18624.0 (calcd mass of 18626.71). The NMR sample was prepared by concentrating the protein to B

DOI: 10.1021/acs.biochem.6b00313 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

atoms with a 15N{1H} NOE higher than 0.6 (residues 33−41, 51−79, 95−123, and 131−174) and were selected to describe the solution structure of rLptE. For further details of assignment and structure calculation, see Table 1.

T1 and T2 relaxation data were measured at both 600 and 700 MHz using [2H,15N]rLptE and HSQC-modified version of the inversion−recovery and CPMG spin echo sequences.25 Relaxation delays were in the range of 0.01−3 s for T1 and 10−300 ms for T2. The 15N{1H} NOE was a proton-detected version of the steady-state experiment.26 T1 and T2 pulse experiments were modified to account for possible solvent exchange as described previously.27 All data were recorded at proton frequencies of 600 and 700 MHz. NMR experiments were taken from the Bruker pulse program library and incorporated whenever possible the Kay− Rance sensitivity enhancement schemes and coherence selections using pulsed-field gradients.28 Proton chemical shifts were calibrated from the positions of the water line at 4.68 ppm at 305 K, from which the carbon and nitrogen scales were derived indirectly using conversion factors of 0.25144954 and 0.10132900, respectively. Spectra were processed in Topspin 2.1 and evaluated in CARA. Peak volumes from relaxation experiments were integrated within SPSCAN that uses lineshape deconvolution of signals for proper integration of partially overlapping peaks and evaluated using home-written routines for least-squares fitting to decaying exponentials. Structure Calculation and Refinement. The solution structure was determined using distance restraints derived from 3D 13C- and 15N-resolved NOESY spectra, both recorded with mixing times of 70 ms. The initial input data set comprised the amino acid sequence, the chemical shift lists, a 3D 15N-resolved 1 H−1H NOESY spectrum, and two 3D 13C-resolved 1H−1H NOESY spectra optimized for the aromatic and aliphatic 13CHn moieties. In total, 1913 NOE upper distance restraints distributed throughout the protein sequence were applied for the final structure calculations. Furthermore, to determine the relative helix and domain orientation in the protein fold, 275 backbone NH, CαC′, and HNC′ RDC restraints were employed during the structure calculations. Initial structures were computed in CYANA29 using chemical shift assignments and unassigned peak lists as input, and annotating peaks with the noeassign macro of CYANA, which in an iterative fashion matches peaks to known chemical shifts and tries to resolve ambiguous assignments using network anchoring. To help with convergence, a homology model based on S. f lexneri LptE [Protein Data Bank (PDB) entry 4Q35; 18% sequence identity] was supplied in the first noeassign cycle. The target function of the final outcome was below 1.0, and the backbone rmsd when superimposing heavy atoms of secondary structure elements was ≈1.2 Å. The structure was then refined by incorporating NH, CαC′, and HC′ RDCs of all residues for which the 15N{1H} NOE was higher than 0.6 to exclude averaging. During the simulated annealing procedure, dihedral angle restraints were applied only during the first two periods of the CYANA protocol, whereas only RDC-derived but no dihedral restraints were applied during the final two steps. When RDC-derived restraints were incorporated, the rmsd decreased to 0.4 Å. Finally, the 20 lowest CYANA conformers were refined in explicit water using the CHARMM22 parallhdg5.3 force field as implemented in XPLOR-NIH.30 The scripts from the xplor-nih-tutorial-2014/ gb1_rdc distribution were utilized for refinement. A total of 1913 NOE-derived distance restraints, 85 NH RDCs, 93 CαCO RDCs, and 236 backbone dihedral angle restraints from TALOS+ were used in this water refinement step. The total energies after XPLOR refinement were −3356 ± 145 kcal/mol. The resulting refined structures were superimposed on all

Table 1. Statistics from the NMR Structure Calculations for rLptE input data for structure calculation no. of NOE distance restraints total intraresidue (i − j = 0) sequential (i − j = 1) medium-range (1 < i − j < 5) long-range (i − j ≥ 5) no. of torsion angle constraints no. of residual dipolar couplings structure statistics, 20 conformers CYANA target function value (Å2) energy (kcal/mol) total distance restraints RDCs torsion angle restraints satisfaction of experimental constraints NOE distance restraints violations of >0.5 Å per structure rmsd of violations RDC restraints violations of >1.5 Hz per structure rmsd of violations torsion angle restraints violations of >5° per structure rmsd of violations PROCHECK Ramachandran plot analysis (%) residues in favored regions residues in additional allowed regions residues in generously allowed regions residues in disallowed regions rmsd from the average coordinates (Å) backbone atoms (residues 33−174) backbone atoms (regular secondary structure)a heavy atoms (residues 33−174) heavy atoms (regular secondary structure)a a

1913 521 617 300 475 252 275 11−12.5 −3356 ± 145 87 ± 13 383 ± 24 14 ± 10

1.2 ± 0.9 0.039 ± 0.003 18.9 ± 1.9 0.86 ± 0.027 1.8 ± 1.6 0.94 ± 0.29 89.2 9.5 0.6 0.7 1.05 0.40 1.60 0.96

± ± ± ±

0.60 0.05 0.52 0.07

Residues 33−41, 51−79, 95−123, and 131−174.

The coordinates of LptE and the chemical shifts, relaxation data, and RDCs were deposited in the PDB and BMRB database as entries 2n8x and 25869, respectively. LPS Binding Studies. A 20 mM stock solution of LPS was prepared in a buffer identical to that used for structural studies of rLptE. LPS (1−10 equiv) was titrated into a 0.2 mM solution of [13C,15N]rLptE, and chemical shift perturbations were followed in [15N,1H]- and [13C,1H]-HSQC spectra.

■

RESULTS Protein Production. Full-length P. aeruginosa lptE (PA3988) is predicted to encode a protein of 207 residues, including a 19-residue signal sequence. After addition of diacylglycerol to the conserved Cys20 residue, signal peptide cleavage, and N-acylation by the Lnt transacylase, Pa LptE will be produced with an N-terminal lipid anchor (e.g., tripalmitoyl, C

DOI: 10.1021/acs.biochem.6b00313 Biochemistry XXXX, XXX, XXX−XXX

Article

Biochemistry

regions. A summary of the structure calculation statistics is given in Table 1. The overall rLptE fold adopts an α/β structure consisting of three α-helices and five β-strands. Three β-strands form an extended antiparallel β-sheet supported by numerous crossstrand backbone NOEs, namely, HN−HN, Hα−Hα, and Hα− HN (Figure S3). Furthermore, the C-terminal segment of β3 forms a parallel β-sheet with β1 and β2, thereby establishing a five-stranded β-sheet. Residues 81−93 are located in the large loop between β3 and β4, which is more flexible (Figure 3). Helix α2 is located on one face of the β-sheet. The residues of the C-terminal part of the α2 helix comprising L148, I149, L152, and L156 form a hydrophobic cluster (core) with I37, L39, β3 L71, L73, β4 I100, Y104, and β5 V117 of strand β1 and V49 and L56 of helix α1 (Figure 3). The core is well-supported by multiple long-range NOEs, and stacking with aromatic side chains results in significant high-field shifts of the involved methyl groups of I37, L56, L73, L152, and L156. The long C-terminal α3 helix (residues 162−178) of rLptE represents a novel feature within this protein family and forms an angle of approximately 120° with respect to helix α2, making contacts with loop residues 107−113 between β-strands β4 and β5. Furthermore, L164 is involved in aliphatic−aromatic stacking with Y69 and I104, which was confirmed by the high-field shift of proton resonances of the methyl groups (δ1/ δ2 CH3). Relaxation Data. The location of protein segments with regular secondary structure correlates well with the 15N relaxation data (see Figure 4). The 15N{1H} NOEs for residues 21−32 are below 0.5, indicating that these residues are not well ordered, and the same is observed for amide moieties from residue 177 on. In addition, 15N{1H} NOEs are reduced for residues 82−94 and 126−135, corresponding to two long loops connecting strands of the β-sheet. Otherwise, the magnitudes of the 15N{1H} NOEs are compatible with a rigidly folded protein. The locations of increased mobility described above also clearly correlate with segments showing reduced R2 rates (Figure 4). Two regions of the protein are characterized by higher-thanaverage R2 values, indicating the presence of exchange effects. One corresponds to residues 38−57, which forms a loop and a short helix that packs against the β-sheet. In addition, residues 135−169 forming the long and kinked C-terminal helix display increased R2 values. Interestingly, a periodicity for enhanced values is seen, indicating that the exchange process affects one side of the helix more than the other (Figure 4). The R2/R1 ratio indicates an overall correlation time of 14.5 ns, which at 305 K is compatible with the presence of a dimer. Line broadening was observed for a large number of residues corresponding to the terminal β3 strand (L73−S88) and the linker between β1 and α1 (S40, Y45, and T48). The same peaks in experiments measured at a concentration of 50 μM were narrow, indicating that the effect arises from selfassociation of rLptE (see Figure S6). Because most of the line broadening occurs for peaks from residues in strand β3, it is likely that strand β3 is part of the dimer interface. A dimer model was made through backbone superposition of regular secondary structure elements of single rLptE conformers with both protomers of the Nm LptE homodimer (PDB entry 3BF2, unpublished work) (Figure S6). This model would also explain well the peak broadening observed in the loop between strand β1 and helix α1, which seems to be additionally involved in dimer formation.

Pam3Cys) before being transferred to the OM along the Lol pathway.31,32 For initial NMR studies, a soluble protein was produced in E. coli comprising LptE residues Gly21−Pro207, with an Nterminal His6 tag, which was removed with the TEV protease, to leave only three additional N-terminal residues (GlyThrSer). However, after large portions of the [ 15N, 1H]-HSQC correlation map of this protein (called here fullLptE) had been assigned, 15N{1H} NOE data revealed that the C-terminal segment of ∼14 residues was largely unstructured and contained mixtures of cis and trans rotamers at one or more Xaa-Pro bonds. A shorter 166-residue protein (Gly21− Arg183), in which all residues with 15N{1H} NOEs of